DataMuseum.dk

top - metrics - download

    Length: 3497 (0xda9)
    Types: TextFile
    Names: »CANONICAL«

└─⟦9ae75bfbd⟧ Bits:30007242 EUUGD3: Starter Kit
    └─⟦bfebc70e2⟧ »EurOpenD3/mail/sendmail-5.65b+IDA-1.4.3.tar.Z« 
        └─⟦f9e35cd84⟧ 
            └─⟦this⟧ »sendmail/ida/cf/CANONICAL« 

TextFile

The internal canonical form used in the IDA/NIU Sendmail.mc.
-----------------------------------------------------------

  The internal representation of an address is a modified source route,
changed so as to provide a completely linear route.

  Thus a source route <@a,@b:user@c>
is converted to:
			@a,@b,@c,user

Notice how the modified source route is a completely linear description
of the sequence of delivery steps.

  We further extend this format by allowing the separation character to be
any of ',' ':' or '!'.  The ',' is used for domains delimited by an '@'.  The
':' is used as a separator for domains delimited by a '%'.  The '!' is the
separator for domains delimited by a '!'.

  Thus an address:  'a!b!c!user'  becomes  '@a!@b!@c!user', while address
'user%b@a' becomes '@a,@b:user'.

  The address is then further canonicalized by enclosing the immediate
domain in <angle brackets>.  Thus a hybrid address 'b!c!x@a' for which the '@'
has precedence is ultimately formatted into: '<@a>,@b!@c!x'.  (However the
initial '@' in the angle brackets may on occasion be changed to a '%' as
a flag which indicates that this is a sender address rather than a recipient
address).

  ALGORITHMS.
  ----------

  The basic idea is that the output address format can be completely set
while retaining the internal canonical form, by merely changing the 
delimiters ',', ':', or '!'.  The address need not be turned back into
external form until the final pass through ruleset 4 (except when used
as a dbm search argument).

  Classes:  CX is '@' and '%' as before.  But CY is changed to '!',',' or ':'.
	    CA is '@' in a class by itself as before.

  Searching for a domain in the address uses the argument:
		@$+$=Y
  Searching for the user portion uses the search argument:
		$=Y$~A$*  (with the last two matching the user).

  Canonicalization is done in ruleset #7 (and its subroutine #9) called
from ruleset #3.  Ruleset #9 is designed to recognize the user portion
of an address.  However in a mixed @/%/! address, the user portion may
itself be a %-path or a !-path.  So rule set #9 canonicalizes any path
in the user portion, which it encloses in <angle brackets> to separate
it from the rest of the address.  Rule set #7 calls #9 to select the
<user portion>, then canonicalizes the address treating the <user portion>
as a unit.  Then it removes the <angle brackets> to merge the possible
two routes.  Finally it places <angle brackets> around the immediate
host.  A local user address (without host) is represented as '<>,user'
to indicate a null host.  (Rule set #3 later changes a local user
to '<>,@local-host-name,user' to provide a default host name).  The
ability to ruleset #7 to combine two routes is also used in the
pathalias lookup (ruleset #22).

  Ruleset #4, which converts an internal address to external form, is
in principle the reverse of the above.  Initially the complete address
is enclosed in <angle brackets>.  Then the idea is to logically look at
what is in <angle brackets> and change it by pulling out the primary
domain in the appropriate external format.  This recursive procedure doesn't
quite work because of the complexity of the source route format, but still
forms the idea behind ruleset #4.
  The unqualification of UUCP nodes with uucpxtable is done in ruleset #4
as part of the formatting process.  This allows the full domain named to
be retained internally for as long as possible.  (However decnetxtable is
not handled in ruleset #4).